Motorola docs tell you how many clock cycles each instruction takes. Don’t forget to adjust for effective address calculation time, if any.
So yeah, I would do it by turning off interrupts, starting a VIA timer, then seeing how long it takes me to do 32767 DBFs. Might be a fun little project actually. Is it possible to turn off instruction caching on an 040? If not you could run it as a totally unrolled loop probably … like literally a thousand or so DBFs in a row.